which setting causes 5241E-Z12E blur with even slow motion?

[carteriii]

I am still fighting with this camera to get readable plates when the vehicle is moving even 1mph. I will factory reset the camera but don't want to do that until the weekend when I have a few hours (in case something goes wrong). In the mean time, would a few of you share which which firmware version you are on if you are using the exact same model (IPC-HFW5241E-Z12E)?

I think you all get the ideal, but this is another video from last night. I don't know if it's a clue or not, but the plate is somewhat more readable as the car is moving forward at the very end of the video. However, it is completely unreadable before the car stops (to enter a code at a community gate), then becomes nice & crisp when the car is completely stopped, then becomes completely unreadable again the moment the car starts to move again, but becomes more readable as the car continues to move forward.

 

[wittaj]

I think there is a wonky setting that a factory reset will fix. It is holding a high gain or high WDR it looks like.

 

[samplenhold]

Still looks like bit rate to me.

I am on

 

[MikeLud1]

Below is the firmware I am on

 

[Flintstone61]

 

[Flintstone61]

this mornings captures. not youtubed,....sorry, i'm rolling out of this joint.....

 

[Flintstone61]

 

[wittaj]

I don't update if my camera is working for me, so I'm older than all of ya LOL

 

[carteriii]

Hey everyone, I'm back after having made many, many, many changes and I want to share some of what I have learned which I believe is NOT intuitive. I'm still not 100% happy so I continue to try different things, but I've learned a lot.

In summary, sharpness, saturation, and iris settings all have a dramatic impact on motion blur in low-light situations with license plates that are a small part of the overall image. When I say, "motion blur", please note that I am not talking about the result of classic slow shutter speed. I've been using 1/2000th of a second (or faster) and I'm talking about the "blur" that I showed at 0:18 in the YouTube video in post #15 of this thread, with a vehicle moving no faster than a human can walk.

Sharpness as a contributor is something that should be somewhat obvious, and I credit @wittaj for sharing his settings with a sharpness=2. I've found that sharpness=0 works even better for me much of the time (but not all!). The Dahua wiki documentation on settings also clearly states that sharpness, "can blur or define details in an image", which I believe shouldn't surprise anyone who understands how that works.

Having said that, here is the next one which I didn't expect. That same Dahua wiki page on settings says, "Also if the picture is over-saturated then it will also cause blurry or artifacting pixels." As I showed in my very first post, I left the saturation alone with the default value of 50, but I've tested and have significantly reduced my "motion blur" with a saturation setting of only 20. I want to do some more exhaustive testing to determine whether I'd be better off with a different value, but my testing shows a dramatic reduction in motion blur with a value of 20 (vs. 50). At a minimum, I did not expect a default saturation setting, on an image which is overwhelmingly black, to have such a negative impact on my motion blur.

Now for the final one, which completely caught me by surprise. A high iris setting creates blur. I'm an old photography guy, very familiar with manually setting an aperture on still/non-video cameras. I've always known the iris to be a physical restriction that controls the amount of light, but I now believe that Dahua also treats the iris setting a bit like "brightness" or "gain", using software/firmware to further adjust the image. In "old school" photography, the primary downside of a wide-open iris was that the depth of field would be reduced, but I've found that with this Dahua camera, a high iris setting contributes to my "motion blur" in low-light situations at distance.

For what it's worth, I discovered the problem with the high iris setting because during the day, I was using shutter priority with otherwise default picture settings. I wanted to ensure a fast shutter speed, thinking I didn't care about the iris/aperture. All my LPR images during the core of daylight were fantastic, but I noticed that around sunrise or sunset, when the camera just flipped to day/color mode or just before the camera would flip over to night (B&W) mode, I would see a similar license plate blur with motion. All the other settings were fixed, so I assumed the cause was a combination of less light and the auto-changing iris.

I want to also mention that when the license plates are larger or brighter, the motion blur is not nearly as significant or at least not enough to prevent reading the plate. Instead, the edges of the license plates digits may be fuzzy, but still readable. But when the plates are further away, so each digit is smaller and the light is less, the same "motion blur" makes it unreadable. If the license plates in your images are simply larger, you may not even realize that you have some of this motion blur occurring, thinking only that your image might be a little "soft". FYI, I'm trying to capture plates 175-200 feet out. This wasn't a critical issue at 150 feet, so I believe I'm also approaching or at the limits of this camera.

With that in mind, I did buy one of the led "laser" lights off AliExpress and it's very nice (for $37 US). It allowed me to experiment with even faster shutter speeds to ensure even 1/2000th of a second was absolutely not the root cause of what I was seeing. There are other threads in these forums about the use of such lights, and it definitely makes a difference, but they have a very, very strong red glow which does draw people's attention to them, unlike the camera's built-in LEDs which have only faint glow.

I continue to tweak/test, and eventually would love to convince my wife to drive back & forth numerous times while I change settings to capture "before & after" clips with each setting. My testing definitely hasn't been scientific, as I'd often change multiple settings at once, out of frustration, but I am confident that individual reductions in sharpness, saturation, and iris have each significantly reduced my "motion blur" at distance with low light.

 

[samplenhold]

In "old school" photography, the primary downside of a wide-open iris was that the depth of field would be reduced, but I've found that with this Dahua camera, a high iris setting contributes to my "motion blur" in low-light situations at distance.

Thanks for your update. I am going to try your recommendations. I've been using sharpness=2 and saturation=50. Even though I am quite happy with my video, one can always improve.

So what do you mean by 'high iris setting'? What number are you using that works for you? Mine is currently set at 50.

 

[carteriii]

I'm currently using iris=30. I was originally using 70-80, thinking that I was only allowing "in" more light.

I don't know how Dahua manages the "iris", or if there even is a physical iris. It would seem there should be, so pure speculation on my part is that the physical iris operates with iris values like (again, just guessing here) 0-50, and then iris values of 51-100 might cause some software changes like brightness & gain that really just manipulate some voltage, current, or setting on the sensor itself, or do some post-process image manipulation. As I said, that's all pure speculation on my part, but I am absolutely confident that iris values of 70-80 cause some image manipulation to happen. There is definitely something more going on than simply letting more light reach the sensor.

 

[Photon Farmer]

I also found numerically high iris settings do affect focus oddly with this model camera.
I mentioned back in post #9 large iris short and long distance were in focus with mid distance at center of capture out of focus.
Large iris reduces depth of field so would expect to lose long or short edge of focus not middle distance.

Gut feeling at this point is you have a networking issue.
Two possible tests, one easy, one not so easy

Easy test:
Try dropping from your 25 FPS to 8 FPS or less.
This would require much less bandwidth.

Not so easy test:
In a different posts you say camera is anywhere from 135 to 200 feet from plates.
I don't see any mention of length of ethernet cable from computer to camera covering community gate
I have learned to keep all runs down to 300 feet
If you have a long run, if possible temporally move computer down to camera and use a short pre made ethernet cable
If you already have a short distance, try a different ethernet cable.

 

[carteriii]

Gut feeling at this point is you have a networking issue.

I am up for testing anything, but I don't think I have a network issue simply because a network problem would cause packet loss or a broken video stream, rather than "corruption" or blurring in only a selective part of my image, particularly during the day when the rest of the image isn't just black. Having said that, I want to be open minded and will definitely try a few of your tests.

I just did some more testing (got my wife to humor me, driving forward & back about 20 times while we talked via the cell phone and I tried different settings), and still found a bit too much "motion blur" for my liking with iris=30 (the setting I mentioned in a previous message). iris=20 was great, but it was too dark for me at the distance I'm trying to get, so I've compromised using iris=25. With the LED/laser light from AliExpress (linked in an earlier message) I'm now able to read plates over 200 feet away (with 100% zoom), which for me is effectively impossible with only the LEDs on the 5241E-Z12E (manually pegged at 100).

In this latest round of tests, I also saw an improvement in my "motion blur" with both sharpness=0 and saturation=0, so for night time settings, that's what I'm going with for now. My "Advanced 3D" noise reduction is set to 40, but the Advanced 2D setting is set to zero.

I do want to share one more super, super useful technique to help yourself at night. This has been mentioned in other posts in the forum, but I think it deserves some repetition and hopefully having it in this same thread will help others.

First, do all your focusing with the focus "speed" set to 1. Even though the UI labels this as focus "speed", I interpret it more as the focus "increment". Night time with limited light requires a far more precise focus, and you need to test smaller increments, and at different distances since the depth of field is not going to be as large as it is during full sunlight. Sometimes, the increments are so small that you can't even see the difference with your eye, even zoomed in. For this, to know what setting you have and to get back to it later, you need to use the API. Use a browser and substitute your own username, password, and IP address in the following URL to get the precise zoom & focus levels.

http://username:password@192.168.1.XX/cgi-bin/devVideoInput.cgi?action=getFocusStatus

That in turn will return output such as:

status.Focus=0.194444
status.FocusMotorSteps=1800
status.LenAdjustStatus=1
status.ResetResult=Success
status.Status=Normal
status.Zoom=1.000000
status.ZoomMotorSteps=1202

Later, when you want to return to a specific zoom and focus, use the API command (via a simple browser URL):

http://username:password@192.168.1.XX/cgi-bin/devVideoInput.cgi?action=adjustFocus&focus=0.194444&zoom=1.000000

Doing that in a browser will simply return an "ok" if you've done it right. I have a scratchpad on the side with various focus & zoom positions that make it quick & easy to test.

 

[bigredfish]

Have been following this and a couple of things come to mind. I’ve setup 6-7 of these over the last 5-6 years and am by no means an expert, but they all follow similar settings.

1- though some have had success at ranges over say 175ft, unless you’ve got ideal conditions, very low angle of attack, and gobs of IR, you’re pushing the limits of the optics of this camera. Onboard IR starts dropping off noticeably at something between 125-135. Sure you can get a good image now and then, but not consistently.

2- Simple is usually better. Once you go down a rabbit hole of tweaking setting on top of setting, it becomes impossible to isolate which one makes the difference. Too many competing combinations.

3- like with any camera/scene it’s all about light. IR or white, but if you don’t have enough, you’re fighting a losing battle. This in my experience is #1 (assuming a decent angle of capture)

4- after light comes focus. Try as I might, many times, the absolute surest way to insure good night focus is either to
A) run b&w 24/7 or
B) switch from color to B&W a min of 30-45 min before sunset.

5- For nighttime LPR, I run 8192 CBR bitrate and 30 fps. More frames gives you more chances of a clear still. With a typical zoomed in scene, youre still covering 20+ feet of capture point. The image focus will change across those frames.

6- some plates just suck

7- many other factors like no backlight, minimal DNR, fixed exposure, lowering sharpness (5241’s are notoriously over sharp) and a small range of Iris adjustment 45-60

My current settings on 2 cameras

5241E-Z12E which setting would you change?

I don't get a bunch of traffic down my street at night and I've been tweaking this camera for a bit now but just can't get it perfect. I am getting some "captures" at night. They all seem readable but deepstack isn't agreeing with that statement. (For the example attached Plate is center of the...

ipcamtalk.com

 

[biggen]

To piggy back on what @bigredfish said, I found that out to 115+', I had to use supplemental IR as the onboard camera just didn't have enough IR to light up some plates very well. I was getting very poor readability for lots of plates at that range. I documented this in my Axton review thread here: Axton PoE Illuminator review. Once I moved to the Axton unit, the only way I miss a plate now is if its covered up or just missing.

 

[carteriii]

I agree with both @bigredfish and @biggen. @biggen in fact your posting on the Axton is one of the things that moved me to buy the illuminator on AliExpress, wanting to first try something less expensive. Thank you for that post.

I do recognize that everything I'm doing/trying is "at the limits", and not for someone who doesn't have a basic understanding of light & image processing, or someone who doesn't have patience. I thought my years of still photography, manually setting shutter speeds & f-stops, would help me, but in fact it may have hurt me. I'm still frustrated that the iris setting is not simply letting in more light by controlling the aperture, and (at least sometimes) is doing some image processing of its own, but I'm glad that I've now learned that. Dahua's documentation might be considered reasonable for the average user, but anyone doing LPR is (or should be) a power user, and the Dahua documentation is miserable for power users. For example, I'm not looking for any trade secrets, but it would be nice to understand why 2D noise reduction made my problem worse, while 3D noise reduction was still necessary to get a readable image.

I have learned a ton from doing this, and I hope it's helpful to everyone else. I think the underlying issue for everyone to appreciate (as @bigredfish called out) is that at night and at distance (aka "at the limits") the plate background isn't a brilliant white or reflective color, and the characters/digits are not already a dark contrast against that white/light/reflective background. And yes, I realize the opposite is true for some plates with a naturally dark background and lighter color characters, but I'm going to use the more traditional light background example for my observations here.

During the day, with a very clear distinction between the plate background and the characters, all the normal image processing (sharpness, saturation, contrast, etc.) works great because the distinction between background and character is already clear. But at night, when the image processing is looking at a plate background that is one shade of grey, and the digits are a second (hopefully much darker or lighter) shade a grey, it's not so easy.

This is compounded by distance when the number of pixels is dramatically less. Searching the web will find statements that LPR needs/deserves at least 100-150 pixels of width to work well. That's a function of your camera resolution, amount of zoom, and proximity to the license plate itself. 150 pixel wide plate would allow the width of each line of each character (i.e. the thickness of the font) to be 5 or more pixels wide (e.g. the width of the number "1"). Think of this like using a "bold" font that makes each line thicker. But at distance, the width of the line of each character might only be 1 or 2 pixels wide, and now they are against a darker "gray" background. Even at 1/2000th of a second, it's not unreasonable that a character has moved just 1 pixel, which is a problem if it was originally only 1 pixel wide. Now there may not be enough pixels and the color difference between the pixels is not as great due to the lack of light. It's easy to understand that standard image processing won't produce the intended results, like saturation or sharpness. I happen to be testing with plates that are less than 100 pixels wide, and I know that, and my expectations are set accordingly. I get readable images during the day with a plate that is only 80 pixels wide, but a night, it's a completely different beast.

Understanding all this is why we cannot simply copy the settings of any one person and have it work perfectly for us in every situation. For example, when I was testing with plate images that were closer to the camera, the plates were were both larger (with thicker digits) and there was more IR light reflecting off the background of the plate, I simply didn't see the motion blur, and in that situation I preferred to have higher settings for things like sharpness, But with smaller images (thinner digits) and less light, my "motion blur" was so bad that I couldn't read the characters. Consequently I got better results with a sharpness setting of zero. With a 50-foot change in distance, or a change in zoom level, I wanted/needed different settings.

I learned a lot and have better results now by standing on the shoulders of all of you who led before me, and shared your experiences. Hopefully this thread and my observations compliment everything the rest of you have done. Thank you all for getting me this far.

 

[Photon Farmer]

Camara networking issues, packets are not the only thing you can drop.

Video from your post #15 reminded me of a problem I had with a HFW5241E-Z12E.
Hikvision NVR manual said check this box and you can run ethernet cable to 1000 feet
Cool, checked box and roughly 750 feet of UV solid copper cat 6 cable later had a little bullet camara
grabbing overview pics, followed about a week later by freshly shipped Z12E from Andy.
Dialed Z12E in and was rewarded with nice license plate pics at about 170 feet from camara
Aghhh... daytime only, nighttime infrared no matter what i did stationary plates could read, however any moment and blurring ensued.
Shutter speed, DNR, WDR if it was adjustable i tried it, I failed.
Love this site, have learned so much from other members, but i finally gave up on this problem.
Added more cameras, at about a dozen cameras the 16 port Hikvision started dropping frames.
Wanting more bandwidth/processing power got a NVR rated to 64 channels.
they don't put 64 ports on the back of these things, so now I needed a POE switch.
I was burning thru boxes of cable making runs 100's of feet long.
Figured why not spread out a few unmanaged switches
now all runs are under 300 feet.
Short time later viewing video I noticed the Z12E night recordings were no longer blurred.
Z12E is now about 60 feet from a switch.
I believe the 750 foot network cable run dropped POE little enough that colour
still worked, yet power drop to much for infrared to work correctly.

 

[carteriii]

I believe the 750 foot network cable run dropped POE little enough that colour
still worked, yet power drop to much for infrared to work correctly.

Ah, that's a great observation! Fortunately I run mostly Ubiquiti Unifi equipment so I was able to just take a look. My PoE switch for the camera is only about 60 feet away, and it shows the camera using up to 10.13W of power right now with the LEDs full blast, and I'm not using even half of the PoE power available. I think I'm good there based on what I see, but it didn't even occur to me to check the power until you suggested it. Thank you.

For what it's worth, I do have one Dahua switch (connected to all other Unifi equipment) running other cameras, including one that is 600+ feet away using Dahua's "extended power over ethernet" which they abbreviate ePoE. It's been rock solid. If anyone really needs extra distance and doesn't want something in the middle, the Dahua ePoE switch & cameras have worked well for me.

 

[richardgohth]

1/2000 seconds? nay. no way, u thinks so. it is not so.

 

[samplenhold]

1/2000 seconds? nay. no way, u thinks so. it is not so.

What? I run mine at 1/2000 at night.